AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Are We Overconfident in Models and Results for Semi-Supervised 3D Medical Image Segmentation?

Jun Li, Ziwei Qin

ICML 2026 · Label-efficient and robust learning › Semi-supervised learning · approach: Semi-supervised medical segmentation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add the uncertainty branch U(v) to dual-axis assessment (ablation)TCSeg · LA/Pancreas-CT/BraTS2019 meanMean DSC (%): 85.68 -> 86.23Improved
add the confidence branch C(v) to dual-axis assessment (ablation)TCSeg · LA/Pancreas-CT/BraTS2019 meanMean DSC (%): 85.20 -> 86.23Improved
add probability-space calibration (ablation)TCSeg · Three datasets meanMean DSC (%): 85.13 -> 86.23Improved
add image-space structural perturbation (ablation)TCSeg · Three datasets meanMean DSC (%): 84.00 -> 86.23Improved
add feature-space prototype alignment (ablation)TCSeg · Three datasets meanMean DSC (%): 80.09 -> 86.23Improved
add feature-space prototype alignment (ablation)TCSeg · Pancreas-CT 10% labeledDSC (%): 59.28 -> 81.08Improved
add feature-space prototype alignment (ablation)TCSeg · Pancreas-CT 20% labeledDSC (%): 69.23 -> 83.44Improved
add semi-supervised tri-space calibration to supervised baseline (ablation)TCSeg · Three datasets meanMean DSC (%): 72.69 -> 86.23Improved
add feature-space prototypes (ablation)TCSeg · BraTS2019 10% labeledDSC (%): 85.57 -> 85.27Hurt
add probability-space calibration (ablation)TCSeg · BraTS2019 10% labeledDSC (%): 85.63 -> 85.27Hurt
widen confidence band or tolerance (0.01 to 0.10) (ablation)TCSeg · Sensitivity analysisDSCNo clear effect
report last instead of best checkpointTCSeg · LA 10% labeledDSC (%): 90.28 -> 90.73No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy